AI Paper Atlas

Research map / 3D vision / Depth and stereo

Sample4Geo: Hard Negative Sampling For Cross-View Geo-Localisation

Fabian Deuser, Konrad Habel, Norbert Oswald

ICCV 2023 · 3D vision › Depth and stereo · approach: Scene coordinate regression for localization · 240 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add GPS + DSS hard negative sampling over random sampling (ablation)Sample4Geo · VIGOR CROSSR@1: 36.38 -> 61.70Improved
add GPS + DSS hard negative sampling over random sampling (ablation)Sample4Geo · VIGOR SAMER@1: 65.23 -> 77.86Improved
add DSS after GPS sampling (combine both) (ablation)Sample4Geo · VIGOR CROSSR@1: 57.17 -> 61.70Improved
GPS sampling instead of random sampling (ablation)Sample4Geo · CVUSAR@1: 97.83 -> 97.83No clear effect
use symmetric InfoNCE instead of satellite-as-query only (ablation)Sample4Geo · VIGOR SAMER@1: 74.69 -> 77.86Improved
use symmetric InfoNCE instead of street-as-query only (ablation)Sample4Geo · VIGOR SAMER@1: 77.49 -> 77.86Improved
use InfoNCE instead of plain triplet loss (ablation)Sample4Geo · CVUSAR@1: 0.00 -> 98.68Improved
use InfoNCE instead of soft-margin triplet (ablation)Sample4Geo · CVUSAR@1: 91.83 -> 98.68Improved
share weights in one encoder instead of two encoders (ablation)Sample4Geo ConvNeXt-B · VIGOR SAMER@1: 75.81 -> 77.86Improved
use ConvNeXt-B instead of ViT backbone (ablation)Sample4Geo · VIGOR SAMER@1: 69.89 -> 75.81Improved
scale backbone from Base to Large (ablation)Sample4Geo · VIGOR SAMER@1: 77.86 -> 78.27No clear effect
increase neighbours k in DSS from 64 to 128 (ablation)Sample4Geo DSS · VIGOR SAMER@1: 77.86 -> 77.84No clear effect

Later papers that compared with this method

9 later accepted papers report Sample4Geo as a baseline in their results tables; 9 of them report a better number on most of the results they share with it.

1 more not listed.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy